Automatic driving vehicle overtaking control method based on nonlinear switching MPC

Abstract

The application provides an automatic driving vehicle overtaking control method based on a nonlinear switching MPC, and relates to the technical field of automatic driving. The method comprises the following steps: acquiring sensing information and real-time running data of a vehicle to be controlled; the real-time running data comprises the attitude and motion state of the vehicle to be controlled; determining a current driving area according to the real-time running data and a sensing distance; matching a corresponding model prediction control mechanism according to the current driving area; wherein a safety area corresponds to a first control mechanism, and a first cost function is configured based on a trajectory tracking error term and a control input smoothing penalty term; an overtaking area corresponds to a second control mechanism, and a second cost function is configured based on a trajectory tracking error term, a control input smoothing penalty term and a double-layer elliptical potential field obstacle avoidance penalty term; solving optimal control parameters according to the model prediction control mechanism, and generating a vehicle control scheme; and controlling the vehicle to complete overtaking according to the control scheme. The application can improve the driving comfort.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to an autonomous vehicle overtaking control method based on nonlinear switching MPC. Background Technology

[0002] With the continuous iteration of intelligent connected vehicle technology, autonomous vehicles are gradually being deployed in traffic scenarios such as highways and urban roads. Automated overtaking, as a core driving behavior for vehicles to actively avoid obstacles, directly affects road safety and the in-vehicle driving experience. Model Predictive Control (MPC), with its advantages of multi-constraint processing and rolling look-ahead optimization, is widely used in overtaking trajectory tracking and longitudinal and lateral cooperative control of autonomous vehicles.

[0003] Artificial potential field method is often combined with model predictive control to achieve obstacle avoidance because of its small computational cost and fast response speed. Conventional artificial potential field only sets the repulsive force based on the straight distance between vehicles. This repulsive force penalty term is directly input into the cost function of model predictive control to participate in optimization.

[0004] Existing autonomous driving overtaking control schemes use the same model to predict the control cost function and solve for the optimal control parameters throughout the entire process. In obstacle-free cruising scenarios, the continuous addition of obstacle avoidance penalty terms can easily cause frequent small fluctuations in the control quantity. In overtaking scenarios where the vehicle is close to an obstacle vehicle, it is impossible to specifically strengthen the collision avoidance constraint. This results in control jitter during the cruising phase and insufficient collision avoidance constraint during the overtaking phase, leading to drastic changes in trajectory and speed throughout the overtaking process and poor driving comfort. Summary of the Invention

[0005] This invention provides an autonomous vehicle overtaking control method based on nonlinear switching MPC to solve the problems of drastic changes in trajectory and speed during overtaking and poor driving comfort in existing autonomous driving overtaking control methods.

[0006] In a first aspect, embodiments of the present invention provide an autonomous vehicle overtaking control method based on nonlinear switching MPC, comprising: Acquire the perception information and real-time operation data of the vehicle to be controlled; wherein, the perception information includes the perception distance between the vehicle to be controlled and the obstacle vehicle, the position of the obstacle vehicle, the heading of the obstacle vehicle, and the speed of the obstacle vehicle; the real-time operation data includes the attitude and motion state of the vehicle to be controlled; Based on the real-time operating data and the perceived distance, the current driving area is determined; wherein, the current driving area includes a safe zone and an overtaking zone; The model prediction control mechanism is matched according to the current driving area; wherein, the safe zone corresponds to the first control mechanism, the overtaking zone corresponds to the second control mechanism, the first control mechanism is configured with a first cost function based on the trajectory tracking error term and the control input smoothing penalty term, and the second control mechanism is configured with a second cost function based on the trajectory tracking error term, the control input smoothing penalty term, and the double-layer elliptical potential field obstacle avoidance penalty term; The optimal control parameters are solved based on the model predictive control mechanism, and a vehicle control scheme is generated. The vehicle is then controlled to complete overtaking according to the control scheme.

[0007] In one possible implementation, the construction process of the first control mechanism includes: Calculate the trajectory tracking error based on the driving trajectory of the vehicle to be controlled and the preset reference trajectory; Calculate a control input smoothing penalty term based on the control input of the vehicle to be controlled; wherein the control input is the speed and steering angle of the vehicle to be controlled. Construct a stage penalty function term based on trajectory tracking error and control input smoothing penalty term; Construct a terminal penalty function term based on the reference trajectory and vehicle terminal constraints; The first cost function is constructed based on the stage penalty function term and the terminal penalty function term; Using the first cost function as the objective function, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed to obtain the first control mechanism.

[0008] In one possible implementation, the construction process of the second control mechanism includes: Calculate the obstacle avoidance penalty term for the double-layer elliptical potential field based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle. The second cost function is constructed based on the stage penalty function term, the terminal penalty function term, and the double-layer elliptic potential field obstacle avoidance penalty term. Using the second cost function as the optimization objective, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed to obtain the second control mechanism.

[0009] In one possible implementation, the step of calculating the double-layer elliptical potential field obstacle avoidance penalty term based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle includes: Based on the preset adjustable parameters of the major semi-axis, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the inner ellipse distance threshold and the outer ellipse distance threshold. The obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. The attraction potential field is determined based on the intensity parameters of the attraction potential field, the lateral position of the centerline of the target road, and the road width adjustment coefficient. Based on the obstacle potential field and the attraction potential field, the double-layer elliptical potential field is determined as the obstacle avoidance penalty term of the double-layer elliptical potential field.

[0010] In one possible implementation, determining the obstacle potential field based on the upper bound of the potential field function, the cost gradient, the inner elliptical distance threshold, the outer elliptical distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle includes: When the distance between the vehicle to be controlled and the obstacle vehicle is less than the inner ellipse distance threshold, the obstacle potential field is the upper bound of the potential field function; When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the inner ellipse distance threshold and less than the outer ellipse distance threshold, the obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the outer ellipse distance threshold, the obstacle potential field is zero.

[0011] In one possible implementation, the step of calculating the inner ellipse distance threshold and the outer ellipse distance threshold based on the preset adjustable semi-major axis parameter, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle includes: Based on the preset adjustable semi-major axis parameters, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the semi-major axis of the inner ellipse and the semi-major axis of the outer ellipse respectively. Calculate the difference in heading angle between the vehicle to be controlled and the obstacle vehicle based on the heading of the vehicle to be controlled and the heading of the obstacle vehicle. When the heading angle difference is within a preset range, the inner ellipse distance threshold is calculated based on the major semi-axis of the inner ellipse, the preset minor semi-axis of the inner ellipse, and the heading of the vehicle to be controlled and the heading of the obstacle vehicle; when the heading angle difference is not within the preset range, the preset minor semi-axis of the inner ellipse is used as the inner ellipse distance threshold. When the heading angle difference is within a preset range, the outer ellipse distance threshold is calculated based on the major semi-axis of the outer ellipse, the preset minor semi-axis of the outer ellipse, and the heading of the vehicle to be controlled and the heading of the obstacle vehicle; when the heading angle difference is not within the preset range, the preset minor semi-axis of the outer ellipse is used as the outer ellipse distance threshold.

[0012] In one possible implementation, solving for the optimal control parameters based on the model predictive control mechanism includes: Based on the control mechanism corresponding to the current driving area, the real-time state of the vehicle to be controlled is assigned as the initial state of the predicted state sequence, and the real-time state parameters of the optimization problem are updated. The objective function of the optimization problem is minimized at each sampling time to obtain the optimal longitudinal velocity and the optimal yaw rate, which are used as the optimal control parameters.

[0013] In one possible implementation, determining the current driving area based on the real-time operating data and the perceived distance includes: The longitudinal relative distance between the vehicle to be controlled and the obstacle vehicle is determined based on the perceived distance; Based on the obstacle vehicle's position, heading, speed, and the attitude and motion state of the vehicle to be controlled in the real-time operation data, calculate the lateral relative distance between the vehicle to be controlled and the obstacle vehicle. When the longitudinal relative distance is greater than the preset area division distance threshold, or when the longitudinal relative distance is less than or equal to the threshold but the lateral relative distance is greater than the safe lane interval, the current driving area is determined to be a safe zone. When the longitudinal relative distance is less than or equal to the preset area division distance threshold, and the lateral relative distance is less than or equal to the safe lane interval, the current driving area is determined to be an overtaking zone.

[0014] In one possible implementation, the perception information is acquired in real time by at least one of an onboard camera, millimeter-wave radar, and lidar.

[0015] In one possible implementation, the attitude and motion state of the vehicle to be controlled includes the longitudinal position, lateral position, speed, yaw angle, and yaw rate of the vehicle to be controlled.

[0016] This invention provides an autonomous vehicle overtaking control method based on nonlinear switching MPC. First, the driving area of ​​the vehicle to be controlled is divided according to real-time operating data and perception distance, providing a reliable basis for selecting different control mechanisms for each area. Then, the first control mechanism corresponding to the safe zone only executes trajectory tracking and input smoothing constraints, while the second control mechanism corresponding to the overtaking zone introduces a double-layer elliptical potential field obstacle avoidance penalty on top of trajectory tracking and smoothing, achieving differentiated control for each area: in the safe zone where there is no collision risk, the focus is on trajectory tracking accuracy and control smoothness, and the first cost function does not include obstacle avoidance-related penalty terms, avoiding redundant constraints from continuously interfering with the optimization process and inducing control oscillations; in the overtaking zone where there is a collision risk, trajectory tracking, control smoothing, and obstacle avoidance constraints are considered simultaneously. This scheme can distinguish between the safe zone and the overtaking zone to configure differentiated cost functions, ensuring overtaking collision avoidance capabilities while suppressing control disturbances, mitigating drastic fluctuations in vehicle speed and trajectory during overtaking, and improving driving comfort. Finally, the optimal control parameters are obtained by solving the matched model predictive control mechanism, generating a control scheme to drive the vehicle to complete autonomous overtaking. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of the autonomous vehicle overtaking control method based on nonlinear switching MPC provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the division of the safety zone and overtaking zone provided in an embodiment of the present invention; Figure 3 This is provided by the embodiments of the present invention. , and Relationship diagram; Figure 4 This is a global trajectory curve diagram of an autonomous vehicle and an obstacle vehicle provided in an embodiment of the present invention; Figure 5 This is a distance curve diagram between an autonomous vehicle and an obstacle vehicle provided in an embodiment of the present invention; Figure 6 This is a control input curve diagram for an autonomous vehicle provided in an embodiment of the present invention; Figure 7 This is a control input constraint curve diagram for an autonomous vehicle provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the lateral and heading angle errors of an autonomous vehicle provided in an embodiment of the present invention; Figure 9 This is the controller switching signal provided in the embodiments of the present invention. Relationship diagram. Detailed Implementation

[0018] With the rapid development of the global economy and the acceleration of urbanization, the number of cars on the road continues to rise, and traffic congestion and safety issues are becoming increasingly prominent, becoming core bottlenecks restricting the high-quality development of the transportation sector. Autonomous driving technology, as a crucial breakthrough for solving traffic safety and improving traffic efficiency, has emerged and quickly become a research hotspot in the global automotive industry and artificial intelligence field, leading the automotive industry into an era of dual transformation: electrification and intelligentization. Although current autonomous driving technology has achieved basic functions such as lane keeping and following, it still has many limitations in overtaking decision-making and control. Constrained by the sensing distance of onboard sensors, overtaking scenarios require algorithm switching between safe cruise and overtaking control modes. However, existing algorithms lack robust switching strategies and stability assurance mechanisms, making them prone to risks such as control variable jumps and system state oscillations during the switching process. Therefore, designing an effective overtaking control strategy for autonomous vehicles that ensures system stability has significant research value and practical significance.

[0019] Model predictive control (MDC) has been widely used in the lateral control of autonomous vehicles due to its advantages such as predictive optimization mechanism, multivariable cooperative control, and strong robustness. Simultaneously, the MDC framework unifies local path planning and tracking control, addressing path tracking error, control input, and collision risk by optimizing the objective function, thus solving the problem of separation between planning and tracking in traditional methods. To ensure the stability of the nonlinear switching mechanism, dividing the entire trajectory tracking scenario into a safe zone and an obstacle avoidance zone is reasonable. However, for many switching systems in practice, common Lyapunov functions may not exist or may be difficult to find. For nonholonomic systems, there is still considerable room for research on switching MDC with multiple Lyapunov functions for trajectory tracking and obstacle avoidance control.

[0020] The purpose of this invention is to provide an overtaking control method for autonomous vehicles based on nonlinear switching MPC. This method establishes the kinematic equations of the autonomous vehicle using the front wheel center point as a reference point, designs compatibility constraints for speed and angular velocity to ensure the safety of lane changes and turns, and designs a two-level elliptical constraint potential field function considering vehicle speed and heading angle to increase longitudinal safety distance, mitigate speed changes during overtaking, and ensure passenger comfort. The entire overtaking region is divided, and an overtaking controller based on nonlinear switching model predictive control is designed, along with a switching mechanism. An average dwell time method closely related to stability conditions is proposed, and multiple Lyapunov functions are set in different regions to analyze the switching stability throughout the process.

[0021] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] See Figure 1The document illustrates a flowchart of the implementation of the autonomous vehicle overtaking control method based on nonlinear switching MPC provided in an embodiment of the present invention, which is described in detail below: Step S101: Obtain the perception information and real-time operation data of the vehicle to be controlled; wherein, the perception information includes the perception distance between the vehicle to be controlled and the obstacle vehicle, the position of the obstacle vehicle, the heading of the obstacle vehicle, and the speed of the obstacle vehicle; the real-time operation data includes the attitude and motion state of the vehicle to be controlled.

[0023] In some embodiments, the vehicle to be controlled is a self-driving vehicle performing an autonomous overtaking maneuver. The perception information is external environmental information acquired through onboard perception modules.

[0024] In this embodiment, the vehicle's motion state and the obstacle vehicle's dynamic environment information are collected simultaneously, providing a complete real-time data source for subsequent driving area determination and model predictive control optimization.

[0025] Step S102: Determine the current driving area based on real-time operating data and perceived distance; wherein, the current driving area includes the safe zone and the overtaking zone.

[0026] In some embodiments, a safe zone is a driving area where the longitudinal distance between the vehicle to be controlled and the obstacle vehicle is relatively large, or the lateral clearance is sufficient, resulting in a low risk of collision. An overtaking zone is an area where the longitudinal distance between the vehicle to be controlled and the obstacle vehicle is less than a defined threshold, and the lateral clearance is insufficient, presenting a risk of collision.

[0027] In practice, the entire area is divided into safe zones based on the sensing distance of the vehicle-mounted sensors. and overtaking area ,like Figure 2 As shown in the figure, the red circle represents a radius of... The sensor's sensing range is shown in the diagram, with the blue ellipse representing the avoidance zone. The area where it is necessary to avoid an obstacle vehicle is called the overtaking zone. Otherwise, it is a safe zone. .

[0028] In this embodiment, the driving conditions are identified based on the lateral and longitudinal relative situations between the vehicle to be controlled and the obstacle vehicle, and safe zones and overtaking zones are divided to provide a basis for judgment for the corresponding model prediction control mechanism for zone matching.

[0029] Step S103: Match the corresponding model prediction control mechanism according to the current driving area; wherein, the safe zone corresponds to the first control mechanism, and the overtaking zone corresponds to the second control mechanism. The first control mechanism is configured with a first cost function based on the trajectory tracking error term and the control input smoothing penalty term, and the second control mechanism is configured with a second cost function based on the trajectory tracking error term, the control input smoothing penalty term, and the double-layer elliptical potential field obstacle avoidance penalty term.

[0030] In some embodiments, the model predictive control mechanism comprises a first control mechanism for the safe zone and a second control mechanism for the overtaking zone, forming a model predictive control optimization framework with differentiated structures. The first control mechanism is used in the safe zone and configures a first cost function; the first cost function includes a trajectory tracking error term and a control input smoothing penalty term, but does not include an obstacle avoidance penalty term. The second control mechanism is used in the overtaking zone and configures a second cost function; the second cost function, in addition to the trajectory tracking error term and the control input smoothing penalty term, adds a double-layer elliptical potential field obstacle avoidance penalty term. The trajectory tracking error term represents the penalty term for the deviation between the vehicle's actual driving trajectory and the reference trajectory. The control input smoothing penalty term is used to constrain drastic changes in speed and steering angle, ensuring smooth control. The double-layer elliptical potential field obstacle avoidance penalty term is a penalty value calculated based on a double-layer elliptical artificial potential field function, used to quantify the obstacle collision risk and achieve obstacle avoidance constraints.

[0031] In this embodiment, differentiated control based on operating conditions is achieved. The safe zone adopts a first cost function without obstacle avoidance penalty terms to eliminate control disturbances caused by redundant obstacle avoidance constraints, focusing on trajectory tracking accuracy and driving smoothness. The overtaking zone adopts a second cost function with added double-layer elliptical potential field obstacle avoidance penalty terms to simultaneously take into account trajectory tracking, control smoothness and obstacle avoidance requirements.

[0032] Step S104: Solve for the optimal control parameters based on the model predictive control mechanism and generate a vehicle control scheme. Control the vehicle to complete the overtaking maneuver according to the control scheme.

[0033] In some embodiments, the optimal control parameters include the optimal longitudinal velocity and the optimal yaw rate.

[0034] In this embodiment, the optimal control parameters are obtained by solving the model predictive control mechanism after matching, and the control scheme is output to drive the vehicle to complete a safe and smooth autonomous driving overtaking maneuver.

[0035] This invention provides an autonomous vehicle overtaking control method based on nonlinear switching MPC. First, the driving area of ​​the vehicle to be controlled is divided according to real-time operating data and perception distance, providing a reliable basis for selecting different control mechanisms for each area. Then, the first control mechanism corresponding to the safe zone only executes trajectory tracking and input smoothing constraints, while the second control mechanism corresponding to the overtaking zone introduces a double-layer elliptical potential field obstacle avoidance penalty on top of trajectory tracking and smoothing, achieving differentiated control for each area: in the safe zone where there is no collision risk, the focus is on trajectory tracking accuracy and control smoothness, and the first cost function does not include obstacle avoidance-related penalty terms, avoiding redundant constraints from continuously interfering with the optimization process and inducing control oscillations; in the overtaking zone where there is a collision risk, trajectory tracking, control smoothing, and obstacle avoidance constraints are considered simultaneously. This scheme can distinguish between the safe zone and the overtaking zone to configure differentiated cost functions, ensuring overtaking collision avoidance capabilities while suppressing control disturbances, mitigating drastic fluctuations in vehicle speed and trajectory during overtaking, and improving driving comfort. Finally, the optimal control parameters are obtained by solving the matched model predictive control mechanism, generating a control scheme to drive the vehicle to complete autonomous overtaking.

[0036] In one possible implementation, the construction process of the first control mechanism includes: Calculate the trajectory tracking error based on the driving trajectory of the vehicle to be controlled and the preset reference trajectory; Calculate the control input smoothing penalty term based on the control input of the vehicle to be controlled; where the control input is the speed and steering angle of the vehicle to be controlled. Construct a stage penalty function term based on trajectory tracking error and control input smoothing penalty term; Construct a terminal penalty function term based on the reference trajectory and vehicle terminal constraints; Construct the first cost function based on the stage penalty function term and the terminal penalty function term; Using the first cost function as the objective function, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed, resulting in the first control mechanism.

[0037] In some embodiments, the stage penalty function term is the cost penalty term at each time step within the prediction time domain of the model predictive control. The terminal penalty function term is the penalty term at the end position of the prediction time domain of the model predictive control, used to ensure long-term state convergence. The vehicle dynamics state equality constraint is the dynamic equation describing the motion law of the vehicle to be controlled, realizing the recursion of the vehicle's motion state. The vehicle dynamics input boundary constraint is the upper and lower limits of speed and steering angle, preventing the control parameters from exceeding the limits of the vehicle actuators. The robust terminal set constraint ensures that the end state in the prediction time domain falls into the stable feasible region, guaranteeing closed-loop stability. The multi-Lyapunov constraint and the mean dwell time constraint are for multi-controller switching conditions, constraining the controller switching behavior and ensuring the stability of the switching process.

[0038] In the specific implementation process, the wheelbase is taken as... With the center point of the front wheel as the reference point, the kinematic equation of the autonomous vehicle can be described as follows: Its corresponding nominal model is given by the following formula: in, It refers to the actual condition of the vehicle. , , These are real-time attitude information in the geodetic coordinate system, representing the vehicle's x-coordinate, y-coordinate, and heading angle, respectively. It is the control input, including yaw rate. and longitudinal velocity . ,in, This refers to the steering angle of the vehicle's front wheels. It is an external disturbance caused by the nonlinear relationship between the vehicle's lateral and yaw motions and its longitudinal velocity, satisfying... , It is a known constant.

[0039] definition This serves as the reference trajectory in the global coordinate system. , , These represent the vehicle's reference x-coordinate, y-coordinate, and heading angle, respectively. The virtual kinematic model corresponding to the reference trajectory is: in, It is a reference control input. and These are the reference linear velocity and angular velocity. The tracking error is calculated as follows: in, , ,and .definition Let the position error vector of the autonomous vehicle be given, and the position error system be as follows: in, , and It consists of linear velocity error and angular velocity error, and To ensure the safety of autonomous vehicles changing lanes and turning during overtaking, the physical and safety constraints are designed as follows: in, This indicates that reversal is prohibited during forward trajectory tracking, and the longitudinal velocity... Limit to 0 to maximum speed between. It can obtain the front wheel steering angle Restricted to from arrive To ensure both comfort and safety during overtaking, constraints are provided. This is to prevent autonomous vehicles from making high-speed turns. To limit centripetal force, the absolute value of centripetal acceleration is less than... Based on the above constraints, the following set of input constraints is constructed. : In the safe area In this context, the first cost function is as follows: Stage penalty function and terminal penalty function They are respectively: in, For the time of update, It is a weight matrix.

[0040] In this embodiment, trajectory deviation is quantified by calculating trajectory tracking error, and speed and steering abrupt changes are constrained by a smoothing penalty term for control input. A first cost function without obstacle avoidance penalty is constructed by combining stage penalty term and terminal penalty term. The model predictive control optimization problem is constructed by integrating dynamics, input boundary, robust terminal set, multi-Lyapunov and average dwell time constraints to obtain the first control mechanism corresponding to the safe zone. Under safe driving conditions, this mechanism can prioritize trajectory tracking accuracy, eliminate redundant obstacle avoidance constraints, avoid unnecessary control disturbances, and maintain stable vehicle driving.

[0041] In one possible implementation, the construction process of the second control mechanism includes: Calculate the obstacle avoidance penalty term for the double-layer elliptical potential field based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle. Construct a second cost function based on the stage penalty function term, the terminal penalty function term, and the double-layer elliptic potential field obstacle avoidance penalty term; Using the second cost function as the optimization objective, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed, resulting in the second control mechanism.

[0042] In the specific implementation process, in the overtaking area In this context, the cost function is as follows: Among them, the stage penalty function and terminal penalty function for: in, Indicates the prediction time domain, This is the switching time. Within the safe zone. In this context, the cost function includes a stage penalty function. and terminal penalty function In the overtaking area In this context, a potential field function is introduced into the cost function. This is to ensure that autonomous vehicles can smoothly change lanes and overtake other vehicles. , and Relationships such as Figure 3 As shown.

[0043] To complete the overtaking maneuver, the following optimization problem is solved at each sampling time: in, Represents the first item in the predicted state sequence. and They are respectively equal to the state and . It is a radius of Robust terminal set. and X and Y represent the major and minor semi-axes of the elliptic safety constraint, respectively, reflecting the vehicle's safe distances in the longitudinal and lateral directions. X and Y are the lateral and longitudinal coordinates of the autonomous vehicle, respectively. and These represent the horizontal and vertical coordinates of the obstacle vehicle, respectively. These are the adjustment parameters for the ellipse constraint. This represents the absolute safety boundary of the dangerous ellipse.

[0044] A mean residence time method closely related to stability conditions is proposed. Multiple Lyapunov functions are set up in different regions to analyze the switching stability throughout the process. The specific steps are as follows: Definition 1: For any time interval ,design For switching signals In the open interval The number of discontinuous points within. An average dwell time exists. and the upper limit of dithering , so that: Theorem 1: Consider a closed-loop system employing a nonlinear switching MPC strategy, let... and Let these represent the cost functions for the safe zone and the overtaking zone, respectively. If parameters exist... satisfy: Then with sampling time As the error increases, the position error system remains stable.

[0045] Proof: The cost function for the optimization problem consists of continuously differentiable terms. This is in contrast to the penalty function for the safe region. Penalty function for overtaking zone An obstacle avoidance penalty potential field function was set. Therefore, it is only necessary to prove Bounded. Choice As a Lyapunov candidate function, it follows the following rules when an autonomous vehicle performs an overtaking maneuver: From the Lipschitz condition, we can obtain: in, Weight matrix The maximum value, This is a local Lipschitz upper bound. Furthermore, assume that for all... , The following conditions must be met: It can be obtained therefore, The upper bound is derived as follows: in, For positive integers, To represent the weight matrix middle and The maximum value. Let , Therefore, the boundedness condition in the theorem is proved.

[0046] Next, we will prove the diminishing condition in the theorem using the cost function of the safe region. We can obtain: Assume that the autonomous vehicle satisfies the local convergence condition: Integrating the above inequality yields: therefore, The decreasing condition in the theorem holds.

[0047] Select in the safe area Optimal cost function and overtaking zone The cost function is used as a Lyapunov candidate function. and The difference is divided into the following three parts: For Part 1 We can obtain: in, To represent the weight matrix The minimum value. For the second part. We can obtain: in, Weight matrix The maximum value. For the third part. We can obtain: exist In, there exists We can obtain: in, , It is the upper bound of external disturbances. When hour, The following inequalities must be satisfied: According to Holder's inequality: get: Combining the above three inequalities, we get: By choosing appropriate parameters, the above inequality can be made less than 0. The theorem is now complete.

[0048] In this embodiment, a double-layer elliptical potential field obstacle avoidance penalty term is calculated based on the motion information of the vehicle to be controlled and the obstacle vehicle to quantify the obstacle collision risk. The obstacle avoidance penalty term is combined with the stage penalty function term and the terminal penalty function term to construct a second cost function, adding obstacle avoidance constraints while retaining trajectory tracking and control smoothness constraints. Furthermore, dynamic constraints, input boundary constraints, robust terminal set constraints, multi-Lyapunov constraints, and average dwell time constraints are integrated to construct a model predictive control optimization problem, resulting in a second control mechanism corresponding to the overtaking area. This allows the vehicle to simultaneously consider trajectory tracking accuracy, control smoothness, and obstacle avoidance capability under overtaking conditions with collision risk, ensuring the overtaking process is safe and controllable.

[0049] In one possible implementation, a two-layer elliptical potential field obstacle avoidance penalty term is calculated based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, including: Based on the preset adjustable parameters of the major semi-axis, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the inner ellipse distance threshold and the outer ellipse distance threshold. The obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. The attraction potential field is determined based on the intensity parameters of the attraction potential field, the lateral position of the centerline of the target road, and the road width adjustment coefficient. Based on the obstacle potential field and the attraction potential field, a double-layer elliptical potential field is determined as the obstacle avoidance penalty term for the double-layer elliptical potential field.

[0050] In some embodiments, the inner elliptical distance threshold and the outer elliptical distance threshold are the two boundaries of the double-layer elliptical potential field; the inner threshold is the high-risk boundary, and the outer threshold is the risk gradient initiation boundary. The upper bound of the potential field function is the maximum potential field value in the region within the inner elliptical boundary, representing the highest collision risk penalty. The cost gradient is a parameter that controls the rate of decay of the potential field value within the interval from the inner threshold to the outer threshold, achieving a smooth transition of the potential field. The obstacle potential field is the repulsive potential field generated by the obstacle; the closer to the obstacle vehicle, the higher the potential field value, used to constrain the vehicle away from the obstacle vehicle. The attractive potential field is an attractive potential field constructed based on the centerline of the target side road; the intensity parameter controls the magnitude of the attraction, and the road width adjustment coefficient limits the lateral range of the attractive potential field, guiding the vehicle to move closer to the centerline of the target lane and avoiding lane crossing.

[0051] In this embodiment, the inner and outer elliptical distance thresholds are dynamically calculated by combining the motion states of the vehicle and the obstacle vehicle, thereby achieving dynamic adaptive adjustment of the obstacle risk area. The obstacle potential field is constructed in a segmented manner, and the value of the potential field is smoothly and gradually changed by using cost gradients to avoid control disturbances caused by abrupt changes in the potential field. Attractive potential field based on the road centerline is constructed simultaneously to guide the vehicle back to the target trajectory and limit the risk of road boundary crossing. The obstacle potential field and the attractive potential field are superimposed to form a double-layer elliptical potential field as an obstacle avoidance penalty term, which can simultaneously take into account obstacle avoidance, trajectory tracking and road boundary constraints in overtaking conditions, thereby improving the safety and stability of overtaking.

[0052] In one possible implementation, the obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner elliptical distance threshold, the outer elliptical distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle, including: When the distance between the vehicle to be controlled and the obstacle vehicle is less than the inner ellipse distance threshold, the obstacle potential field is the upper bound of the potential field function. When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the inner ellipse distance threshold and less than the outer ellipse distance threshold, the obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the outer ellipse distance threshold, the obstacle potential field is zero.

[0053] In this embodiment, a three-stage rule is used to construct the obstacle potential field: the maximum potential field is maintained within the inner ellipse closest to the obstacle vehicle to strengthen collision avoidance constraints; the obstacle potential field continuously and smoothly decays in the middle interval between the inner and outer ellipses using a cost gradient to eliminate abrupt changes in the potential field; when the distance exceeds the threshold of the outer ellipse, the potential field is set to zero, releasing the obstacle penalty constraint. This method can dynamically adjust the collision penalty intensity according to the relative distance between the vehicle and the obstacle vehicle, ensuring the safety of close-range overtaking collision avoidance while achieving a smooth transition of potential field force, suppressing control abrupt changes, and improving driving comfort.

[0054] In one possible implementation, based on preset adjustable parameters of the major semi-axis, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, the inner ellipse distance threshold and the outer ellipse distance threshold are calculated, including: Based on the preset adjustable parameters of the major semi-axis, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the major semi-axis of the inner ellipse and the major semi-axis of the outer ellipse respectively. Calculate the difference in heading angle between the vehicle to be controlled and the obstacle vehicle based on their headings; When the heading angle difference is within the preset range, the inner ellipse distance threshold is calculated based on the major semi-axis of the inner ellipse, the preset minor semi-axis of the inner ellipse, and the heading of the vehicle to be controlled and the heading of the obstacle vehicle; when the heading angle difference is not within the preset range, the preset minor semi-axis of the inner ellipse is used as the inner ellipse distance threshold. When the heading angle difference is within the preset range, the outer ellipse distance threshold is calculated based on the major semi-axis of the outer ellipse, the preset minor semi-axis of the outer ellipse, and the headings of the vehicle to be controlled and the obstacle vehicle. When the heading angle difference is not within the preset range, the preset minor semi-axis of the outer ellipse is used as the outer ellipse distance threshold.

[0055] In some embodiments, the major semi-axis of the inner ellipse and the major semi-axis of the outer ellipse correspond to the axial lengths of the inner and outer ellipses of the double-layer elliptical potential field along the relative heading direction, respectively. The preset minor semi-axis of the inner ellipse and the preset minor semi-axis of the outer ellipse are the minor dimensions of the inner and outer ellipses of the double-layer elliptical potential field.

[0056] In practical implementation, the obstacle potential field serves to keep the autonomous vehicle away from the obstacle vehicle. Velocity and heading angle are incorporated into the construction of the obstacle potential field function, based on the distance between the autonomous vehicle and the obstacle vehicle. The two-level elliptical barrier potential field function can be divided into the following three piecewise functions: in, This is represented as the upper bound of the penalty function (and potential function). The cost gradient can be controlled within the collision area and edge area. k=0,1 and These represent the inner ellipse distance threshold and the outer ellipse distance threshold, respectively. This is based on the heading angle difference between the autonomous vehicle and the obstacle vehicle. The parameter range can be divided into the following two segments: in, and Let be the major and minor semi-axis of the elliptic potential field, respectively. Defined as: in, This represents the semi-major axis of the inner ellipse. This represents the semi-major axis of the outer ellipse. and The parameter is adjustable for the long half-shaft. It is the heading of the vehicle to be controlled. Indicates the heading of the obstacle vehicle. The speed of the obstacle vehicle.

[0057] To avoid the control oscillations and driving stagnation problems caused by the unreachability of the target in the traditional Artificial Potential Field (APF) method, the attractive potential field of the target lane is described by the following function to guide the autonomous vehicle: in, It is an adjustable intensity parameter of the attractive potential field. It is the vehicle's ordinate in the geodetic coordinate system. Indicates the lateral position of the centerline of the target road. This indicates the road width adjustment coefficient.

[0058] The APF function can be expressed as the sum of the obstacle potential field and the attractive potential field, that is: In this embodiment, the major semi-axis of the inner and outer ellipses is dynamically calculated based on the motion states of the vehicle to be controlled and the obstacle vehicle. The distance threshold solution strategy is selected according to the difference in heading angle between the two vehicles: when the heading difference is within a preset range, the distance threshold between the inner and outer layers is dynamically calculated based on the ellipse parameters; when the heading difference exceeds the preset range, the minor semi-axis of the ellipse is directly used as the threshold. This realizes that the boundary of the double-layer elliptical potential field is dynamically and adaptively adjusted with the relative heading of the two vehicles, which overcomes the defect that the traditional fixed circular obstacle avoidance area is difficult to adapt to the changes in the heading of the two vehicles, improves the rationality of obstacle risk area judgment, and takes into account the overtaking and obstacle avoidance effect under different intersection situations.

[0059] In one possible implementation, the optimal control parameters are solved based on the model predictive control mechanism, including: Based on the control mechanism corresponding to the current driving area, the real-time state of the vehicle to be controlled is assigned as the initial state of the predicted state sequence, and the real-time state parameters of the optimization problem are updated. The objective function of the optimization problem is minimized at each sampling time to obtain the optimal longitudinal velocity and the optimal yaw rate, which are used as the optimal control parameters.

[0060] In this embodiment, the current real-time motion state of the vehicle to be controlled is used as the initial value for predicting the time-domain state recursion. The model parameters are updated and optimized in real time to ensure that the future motion state of the vehicle is closely related to the actual driving conditions. Furthermore, the optimal longitudinal speed and optimal yaw rate are solved by continuously minimizing the cost function of the corresponding control mechanism according to the sampling period, so as to realize the coordinated control of the vehicle's longitudinal driving and lateral overtaking actions, and output real-time and reasonable autonomous driving overtaking control commands.

[0061] In one possible implementation, the current driving area is determined based on real-time operational data and perceived distance, including: The longitudinal relative distance between the vehicle to be controlled and the obstacle vehicle is determined based on the perceived distance; Based on the obstacle vehicle's position, heading, speed, and the attitude and motion state of the vehicle to be controlled in the real-time operation data, calculate the lateral relative distance between the vehicle to be controlled and the obstacle vehicle. When the longitudinal relative distance is greater than the preset area division distance threshold, or when the longitudinal relative distance is less than or equal to the threshold but the lateral relative distance is greater than the safe lane interval, the current driving area is determined to be a safe zone. When the longitudinal relative distance is less than or equal to the preset area division distance threshold, and the lateral relative distance is less than or equal to the safe lane interval, the current driving area is determined to be an overtaking zone.

[0062] In some embodiments, the longitudinal relative distance is the distance between the vehicle and the obstacle vehicle along the road travel direction. The lateral relative distance is the lateral positional relationship obtained by combining the positions of the vehicle and the obstacle vehicle. A preset zone division distance threshold is used to distinguish the longitudinal critical distance between the safe zone and the overtaking zone. The safe lane spacing is the lateral safe lane spacing standard.

[0063] In this embodiment, the driving area is determined by combining the longitudinal relative distance and lateral relative situation between the vehicle and the obstacle vehicle. The driving conditions and overtaking conditions are distinguished by comprehensively considering the longitudinal and lateral two-dimensional relative situations. When the longitudinal distance is sufficient or the lateral avoidance space is sufficient, it is determined to be a safe driving condition, and only the first control mechanism that takes into account both trajectory tracking and control smoothness is used. Only in high-risk scenarios where the longitudinal distance is close and the lateral space is limited is it determined to be an overtaking condition and the second control mechanism with obstacle avoidance penalty with double-layer elliptical potential field is activated. This achieves accurate identification of the driving conditions and provides a reliable judgment benchmark for matching the regional differentiated model predictive control mechanism.

[0064] In one possible implementation, the perception information is acquired in real time through at least one of an onboard camera, millimeter-wave radar, and lidar.

[0065] In one possible implementation, the attitude and motion state of the vehicle to be controlled include the longitudinal position, lateral position, speed, yaw angle, and yaw rate of the vehicle to be controlled.

[0066] This invention establishes the kinematic equations of an autonomous vehicle using the center point of the front wheels as a reference point, and designs compatibility constraints for speed and angular velocity to ensure the safety of lane changes and turns. To increase the longitudinal safety distance, mitigate speed changes during overtaking, and ensure passenger comfort during overtaking, a two-level elliptical constraint potential field function considering vehicle speed and heading angle is designed. The entire overtaking area is divided, and an overtaking controller based on nonlinear switching model predictive control is designed, along with a switching mechanism. An average dwell time method closely related to stability conditions is proposed, and multiple Lyapunov functions are set in different regions to analyze the switching stability throughout the process. Therefore, this invention is more suitable for practical application scenarios.

[0067] Compared with the prior art, the present invention has the following advantages: 1. Considering the local optima and target unreachability problems of traditional artificial potential fields, a two-level elliptical constrained potential field function considering vehicle speed and heading angle is designed, and an attractive potential field is established based on the centerline of the road on the target side. 2. Based on the sensing distance of the onboard sensors of the autonomous vehicle, the entire overtaking area is divided into a safe zone and an overtaking zone, and corresponding model prediction controllers are designed for each zone, and a switching mechanism is established.

[0068] 3. An average residence time method closely related to stability conditions is proposed, and multiple Lyapunov functions are set in different regions to analyze the switching stability throughout the process.

[0069] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0070] To verify the effectiveness of the proposed nonlinear switching MPC-based overtaking control method for autonomous vehicles, experimental results are presented as verification, demonstrating that the method has significant advantages in balancing safety and comfort with ensuring the stability of switching control during the overtaking process of autonomous vehicles. The specific implementation steps of the experiment are as follows: Step S1.1: The experimental platform robot serves as both an autonomous driving vehicle and an obstacle vehicle. It integrates a CPU computing unit, an inertial measurement unit (IMU), a high-precision speed encoder, an ultra-wideband positioning system (UWB), a binocular depth camera, and a lidar. In the overtaking scenario, the autonomous driving vehicle travels along the centerline of the current straight road at a reference speed of 0.25 m / s, while the obstacle vehicle travels in the same direction and lane at a reference speed of 0.1 m / s.

[0071] Step S2.1: Provide experimental parameters: Adjustable parameters in the potential field function... and Set to respectively , The major semi-axis of the elliptical safety distance model and short half shaft Set them to 0.85 and 0.45 respectively.

[0072] Step S3.1: Select the weight matrix in the controller cost function. , and Set them to: Figure 4 , Figure 5 These are the global trajectory of the autonomous vehicle and the distance between it and the obstacle vehicle, respectively. The minimum distance between the two vehicles is 0.6m, which is greater than the safe distance of 0.5m, meaning that there is no collision during the entire overtaking process.

[0073] Figure 6 , Figure 7 The control input and control input constraint index are respectively used. The speed and angular velocity obtained by the controller are predicted using the switching model, and their values ​​are within the amplitude constraint range.

[0074] Figure 8 , Figure 9 These are the lateral and heading angle errors and the controller switching signal, respectively. After the autonomous vehicle completes the overtaking maneuver, the error converges, and the switching control strategy is stable.

[0075] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0076] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for overtaking control of autonomous vehicles based on nonlinear switching MPC, characterized in that, include: Acquire the perception information and real-time operation data of the vehicle to be controlled; wherein, the perception information includes the perception distance between the vehicle to be controlled and the obstacle vehicle, the position of the obstacle vehicle, the heading of the obstacle vehicle, and the speed of the obstacle vehicle; the real-time operation data includes the attitude and motion state of the vehicle to be controlled; Based on the real-time operating data and the perceived distance, the current driving area is determined; wherein, the current driving area includes a safe zone and an overtaking zone; The model prediction control mechanism is matched according to the current driving area; wherein, the safe zone corresponds to the first control mechanism, the overtaking zone corresponds to the second control mechanism, the first control mechanism is configured with a first cost function based on the trajectory tracking error term and the control input smoothing penalty term, and the second control mechanism is configured with a second cost function based on the trajectory tracking error term, the control input smoothing penalty term, and the double-layer elliptical potential field obstacle avoidance penalty term; The optimal control parameters are solved based on the model predictive control mechanism, and a vehicle control scheme is generated. The vehicle is then controlled to complete overtaking according to the control scheme.

2. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, The construction process of the first control mechanism includes: Calculate the trajectory tracking error based on the driving trajectory of the vehicle to be controlled and the preset reference trajectory; Calculate a control input smoothing penalty term based on the control input of the vehicle to be controlled; wherein the control input is the speed and steering angle of the vehicle to be controlled. Construct a stage penalty function term based on trajectory tracking error and control input smoothing penalty term; Construct a terminal penalty function term based on the reference trajectory and vehicle terminal constraints; The first cost function is constructed based on the stage penalty function term and the terminal penalty function term; Using the first cost function as the objective function, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed to obtain the first control mechanism.

3. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, The construction process of the second control mechanism includes: Calculate the obstacle avoidance penalty term for the double-layer elliptical potential field based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle. The second cost function is constructed based on the stage penalty function term, the terminal penalty function term, and the double-layer elliptic potential field obstacle avoidance penalty term. Using the second cost function as the optimization objective, and combining vehicle dynamics state equality constraints, vehicle dynamics input boundary constraints, robust terminal set constraints, multiple Lyapunov constraints, and average dwell time constraints, a model predictive control optimization problem is constructed to obtain the second control mechanism.

4. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 3, characterized in that, The calculation of the double-layer elliptical potential field obstacle avoidance penalty term based on the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle includes: Based on the preset adjustable parameters of the major semi-axis, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the inner ellipse distance threshold and the outer ellipse distance threshold. The obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. The attraction potential field is determined based on the intensity parameters of the attraction potential field, the lateral position of the centerline of the target road, and the road width adjustment coefficient. Based on the obstacle potential field and the attraction potential field, the double-layer elliptical potential field is determined as the obstacle avoidance penalty term of the double-layer elliptical potential field.

5. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 4, characterized in that, The step of determining the obstacle potential field based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle includes: When the distance between the vehicle to be controlled and the obstacle vehicle is less than the inner ellipse distance threshold, the obstacle potential field is the upper bound of the potential field function; When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the inner ellipse distance threshold and less than the outer ellipse distance threshold, the obstacle potential field is determined based on the upper bound of the potential field function, the cost gradient, the inner ellipse distance threshold, the outer ellipse distance threshold, and the distance between the vehicle to be controlled and the obstacle vehicle. When the distance between the vehicle to be controlled and the obstacle vehicle is greater than or equal to the outer ellipse distance threshold, the obstacle potential field is zero.

6. The overtaking control method for autonomous vehicles based on nonlinear switching MPC according to claim 4, characterized in that, The step of calculating the inner ellipse distance threshold and the outer ellipse distance threshold based on the preset adjustable semi-axis parameters, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle includes: Based on the preset adjustable semi-major axis parameters, the speed of the vehicle to be controlled, the speed of the obstacle vehicle, the heading of the vehicle to be controlled, and the heading of the obstacle vehicle, calculate the semi-major axis of the inner ellipse and the semi-major axis of the outer ellipse respectively. Calculate the difference in heading angle between the vehicle to be controlled and the obstacle vehicle based on the heading of the vehicle to be controlled and the heading of the obstacle vehicle. When the heading angle difference is within a preset range, the inner ellipse distance threshold is calculated based on the major semi-axis of the inner ellipse, the preset minor semi-axis of the inner ellipse, and the heading of the vehicle to be controlled and the heading of the obstacle vehicle; when the heading angle difference is not within the preset range, the preset minor semi-axis of the inner ellipse is used as the inner ellipse distance threshold. When the heading angle difference is within a preset range, the outer ellipse distance threshold is calculated based on the major semi-axis of the outer ellipse, the preset minor semi-axis of the outer ellipse, and the heading of the vehicle to be controlled and the heading of the obstacle vehicle; when the heading angle difference is not within the preset range, the preset minor semi-axis of the outer ellipse is used as the outer ellipse distance threshold.

7. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, Solving for the optimal control parameters based on the model predictive control mechanism includes: Based on the control mechanism corresponding to the current driving area, the real-time state of the vehicle to be controlled is assigned as the initial state of the predicted state sequence, and the real-time state parameters of the optimization problem are updated. The objective function of the optimization problem is minimized at each sampling time to obtain the optimal longitudinal velocity and the optimal yaw rate, which are used as the optimal control parameters.

8. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, The step of determining the current driving area based on the real-time operating data and the perceived distance includes: The longitudinal relative distance between the vehicle to be controlled and the obstacle vehicle is determined based on the perceived distance; Based on the obstacle vehicle's position, heading, speed, and the attitude and motion state of the vehicle to be controlled in the real-time operation data, calculate the lateral relative distance between the vehicle to be controlled and the obstacle vehicle. When the longitudinal relative distance is greater than the preset area division distance threshold, or when the longitudinal relative distance is less than or equal to the threshold but the lateral relative distance is greater than the safe lane interval, the current driving area is determined to be a safe zone. When the longitudinal relative distance is less than or equal to the preset area division distance threshold, and the lateral relative distance is less than or equal to the safe lane interval, the current driving area is determined to be an overtaking zone.

9. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, The perception information is acquired in real time through at least one of vehicle-mounted cameras, millimeter-wave radar, and lidar.

10. The autonomous vehicle overtaking control method based on nonlinear switching MPC according to claim 1, characterized in that, The attitude and motion state of the vehicle to be controlled include the longitudinal position, lateral position, speed, yaw angle, and yaw rate of the vehicle to be controlled.

Images